Amputee sprinter Dilip Gavit eyes Asian Games glory after Commonwealth record

Dilip Gavit, a 23-year-old para-athlete from a tribal village in Nashik, lost his right arm in a childhood accident but went on to build a remarkable sporting career. He competed in the 400m before making an unlikely switch to the 100m sprint. His perseverance, guided by a pivotal coach, helped him reach the Commonwealth Games in Glasgow, where he set a record. Gavit now has his sights set on the Asian Games as his next major milestone.

QNu Labs raises Rs 200 crore in funding

Alco-bev startup Marmalade raises Rs 6 crore

Small-tier IT punches above their weight

India on track for 100-plus GCCs in 2026

Flipkart, Amazon's multi-front war

Yash's 'Toxic' Collects Rs 69 Lakh on Day 14, Total Nears Rs 247 Crore

Yash-starrer 'Toxic' brought in Rs 69 lakh on its fourteenth day, marking an 8% dip compared to the previous Tuesday. The film's cumulative India net collection has climbed to Rs 247 crore. Globally, the movie has grossed Rs 338 crore in total. The Kannada version outperformed the Hindi version in day-14 earnings. Overall, the film has seen a notable slowdown in momentum during its second week of release.

Harsimrat Badal Files Complaint Against Bhagwant Mann and Wife With Women's Body

Shiromani Akali Dal leader Harsimrat Kaur Badal has filed a formal complaint with a women's commission against Punjab Chief Minister Bhagwant Mann and his wife. The complaint comes amid escalating political tensions in Punjab between the Akali Dal and the ruling Aam Aadmi Party. Badal alleged that the AAP government has caused a significant decline in Punjab's overall conditions during its five years in power. The move signals a sharp intensification of opposition criticism targeting the Chief Minister and his administration.

ChaosCypher v0.4.2 Patches Streaming Bug, Path Traversal, and Rate-Limit Flaws

ChaosCypher has released v0.4.2, a maintenance update containing 61 bug fixes with no new features or API changes. The most significant fix restores live chat streaming, which had silently failed due to middleware tearing down storage adapters before responses finished sending — the bug was hard to spot because completed answers still appeared on page reload. A second fix closes a path traversal vulnerability where a queued task could create SQLite database files at arbitrary locations via an unsanitised metadata field. Two rate-limiting defects were also corrected: the health endpoint lacked request throttling, and a misconfigured nginx zone caused per-IP login limits to collapse into a single global bucket, posing an availability risk on local networks. Self-hosted users are advised to upgrade, and the previously untested middleware now has seven new test cases covering the fixed behaviour.

'You can't keep it lingering': Supreme Court to Odisha on Dara Singh plea

Why Amazon Dropped MOBI for EPUB and How Browsers Can Convert Files Privately

Amazon formally phased out MOBI support for its Send to Kindle service in late 2022, pushing users toward the EPUB format instead. MOBI, originally created by Mobipocket in 2000 and later acquired by Amazon in 2005, lacked modern typography features such as CSS layouts, SVG support, and dynamic dark mode styling. EPUB, standardized by the W3C and IDPF, is built on HTML5, CSS3, and XML, making it far better suited to contemporary Kindle rendering engines. When users sought alternatives to convert their existing MOBI libraries, many turned to online tools that upload files to third-party servers, raising serious privacy concerns around data retention and exposure. Modern browser technologies like WebAssembly and FileReader APIs now enable fully client-side MOBI-to-EPUB conversion, meaning files never leave the user's device.

How to Deploy a Web Project to Tencent EdgeOne Makers via GitHub

Tencent EdgeOne Makers offers a streamlined way to deploy web projects by connecting directly to a GitHub repository, eliminating the need for manual file uploads or server configuration. Developers can sign in to the EdgeOne Makers dashboard, authorize access to their GitHub account, and select the desired repository for deployment. The platform automatically detects common frameworks like React, Vue, or Vite and suggests appropriate build settings. Once deployed, EdgeOne Makers provides a live URL to access the website. Any subsequent code changes pushed to the connected GitHub branch can automatically trigger a new deployment, keeping the live site up to date.

Building collapses to fires, 33k died in 5 years from 2020

Agent benchmarks measure memory-less models, missing a key performance variable

A critique published on DEV Community argues that AI coding-agent benchmarks are structurally flawed because they always start models with empty memory, ignoring how accumulated project context affects real-world performance. The argument uses Engrim, an open-source SQLite-based persistent memory engine for tools like Claude Code and Cursor, as evidence that cross-session state is a meaningful and underexplored performance axis. Current evals deliberately reset memory between runs to ensure reproducibility, but this means the benchmarked agent bears little resemblance to one that has ingested months of project decisions. The author proposes a fix that preserves reproducibility: run each task under three conditions — empty context, small curated memory, and large memory — and measure how scores shift across conditions. Until benchmarks account for persistent state, their results represent only a performance floor, not a reliable predictor of production behavior.

Why Fluent but Wrong AI Code Reviews Are More Dangerous Than Obvious Errors

A analysis published on DEV Community argues that the most dangerous failure in AI code review tools is not a missed bug but a structurally incorrect review that appears credible and trustworthy. Drawing on a database-recovery case study by Oskar Gross at Glazer, the piece illustrates how readable output with corrupted meaning — like CSV values under wrong headers — is worse than an obvious error because it goes unquestioned. The author warns that many AI review tools assess only whether code reads well, rather than validating it against the broader codebase, its types, contracts, and callers. A tool that produces fluent but structurally blind feedback can cause developers to stop scrutinizing changes precisely where scrutiny is most needed. The piece urges teams to benchmark AI review tools on worst-case scenarios — specifically the subset of changes where the tool gives plausible but wrong feedback — rather than relying on average bug-catch rates.

How Tool Calling Enables AI Agents to Act Beyond Their Training Data

Large language models (LLMs) cannot independently fetch real-time information such as weather updates or stock prices, so they rely on external tools to perform such actions. When a user submits a query, the LLM analyzes it to determine whether any registered tool matches the request, then suggests the appropriate tool along with required parameters. An external logic layer executes the action and returns the result to the LLM, which then generates a structured response for the user. If no matching tool is found, the LLM falls back on its own trained knowledge to answer the query. Accurate and descriptive tool documentation is critical, as vague descriptions can cause the LLM to fail in correctly identifying or invoking the right tool.

SA Developer Builds Free Tool to Centralise University Info for Matric Students

A South African developer has launched Matric Helper, a free web tool designed to simplify university applications for matric students. The platform consolidates information from all 26 South African public universities, including application deadlines, APS requirements, and application fees, into a single location. Built using Flask, Jinja2, and vanilla CSS with no database, the project was created by the developer while still learning Python fundamentals. The tool is aimed particularly at first-generation university applicants who lack guidance navigating multiple university websites. The site is live and deployed on Render, with the source code publicly available on GitHub.

From Imitation to Reasoning: How LLMs Evolved Beyond Simple Instruction Tuning

Major AI models like GPT-4.5, DeepSeek-V3, and Claude 3.5 Sonnet were the last flagship systems built entirely on the traditional pretrain-then-instruct approach, where all computation occurred in a single forward pass with no revision mechanisms. By late 2024, AI labs hit a scaling ceiling where adding more data, parameters, or compute no longer yielded proportional gains, especially on complex multi-step reasoning tasks. This prompted a shift toward training models to 'think before they answer' using reinforcement learning techniques layered on top of existing training stages. Supervised fine-tuning (SFT), which teaches models by imitating human-annotated examples, proved limited because model quality was capped by annotator skill and offered no mechanism for exploring better alternatives. Reinforcement learning methods like RLHF addressed this by allowing models to generate multiple candidate responses and receive feedback signals indicating which outputs were actually superior.

US bans Canadian alcohol imports amid escalating bilateral trade war

The United States has imposed an import ban on Canadian alcohol and other goods, marking a fresh escalation in the ongoing trade dispute between the two nations. The move comes as Canada's retaliatory tariffs on American goods simultaneously came into force. The tit-for-tat measures reflect deepening tensions in cross-border trade relations between the two neighboring countries. Both sides appear to be intensifying economic pressure on each other with no immediate resolution in sight.

Security job scan of 37 freelance posts finds none listed within the past 24 hours

A developer ran an informal audit on a large freelance marketplace on September 8, 2026, searching for security-related jobs using five keyword anchors including 'wazuh', 'splunk', and 'soc 2'. Of 37 job cards found — representing 33 unique postings — not a single one had been listed within the previous 24 hours, with the most recent being a day old. The analysis also uncovered a flaw in the author's own age-parsing script, which returned null for listings labelled 'Posted last week', nearly leading to a false conclusion about platform activity. A separate pattern emerged around buyer intent: the search term 'soc 2' returned a higher proportion of verified, high-spending clients than the four product-name anchors combined. The author notes these are preliminary observations from a single session and invites others selling security or SOC services on freelance platforms to compare their own data.